Hyper-threading technology

Hi

We have just bought a new machine, it is a Dell presicion 360, 3 Ghz,

I did a benchmark test, a Specapc test and it ran 5 times.

Test Averages for 5 tests. Test Total = 394 Graphics = 240 CPU = 83 I/O = 72

Then I enabled Hyper Threading.

Test Averages for 5 test(s). Test Total = 388 Graphics = 241 CPU = 74 I/O = 74

So my little conclusion must that I gain in SolidWorks with Hyper Threading.

If anybody else can make the same test, it could be interesting to see if you can get same result.

I wish you all a merry Christmas and a happy New Year.

Klaus

Hi

We have just bought a new machine, it is a Dell presicion 360, 3 Ghz,

I did a benchmark test, a Specapc test and it ran 5 times.

Test Averages for 5 tests. Test Total = 394 Graphics = 240 CPU = 83 I/O = 72

Then I enabled Hyper Threading.

Test Averages for 5 test(s). Test Total = 388 Graphics = 241 CPU = 74 I/O = 74

So my little conclusion must that I gain in SolidWorks with Hyper Threading.

If anybody else can make the same test, it could be interesting to see if you can get same result.

I wish you all a merry Christmas and a happy New Year.

Klaus

The SPECAPC benchmark is heavily weighted towards graphics. There is no evidence in your numbers that anything noticable changed there.

The CPU and I/O portions of the benchmark are limited and not really representative of real world work habits. You might want to find a large assebmly or long rebuild part and do a comparitive test there before getting too excited about the CPU numbers.

Klaus Sabroe wrote:

Well, I wouldn't have believed it unless I read what Ed posted so I decided to give it a try:

one processor (no hyperthread) = 191 seconds one processor (with hyperthread) = 174 seconds two processors (no hyperthread) = 97 seconds two processors (with hyperthread) 129 seconds

looks like hyperthread does help on one processor but doesn't on dual machines. This isn't really reliable data unless I test same assembly on single processor machine.

The assembly is of a handheld device with metalic, reflective and transparent materials.

PW2 settings: one reflection bounce, Anti-alias= high, max memory allocation = 990mb, no indirect illumination

hardware: Dell dual Xeon 2.66GHz, 2 Gigs RAM, Nvida QuadroFX3000 graphics WinXP Pro

Mark,

I d> Well, I wouldn't have believed it unless I read what Ed posted so I

Yes, your correct, I was simply verifying Ed's comments on PW which takes advantage of mult-threading.

having hyper-threading on while using solidworks is slower! On a duel Monitor system. hyper-threading is faster because you can work in other applications while solidworks rebuilds. while using applications during a rebuild. that application does not feel like its chocking for cpu time. but the solidworks rebuild will be slower. Do you devide your time? or go from rebuild to rebuild? do you devide your cpu? or go from start to finish like a solidworks rebuild?

Mark wasn't benchmarking SolidWorks. He was giving PhotoWorks times. Klaus was using the SPECApc benchmark which is for SolidWorks. It wasn't an apples to apples comparison because PhotoWorks does use multiprocessing and SolidWorks does not.

What Mark did provide was the fact that hyper-threading and multi-processing don't seem to work well together. Whether this is a Windoze thing or just a PhotoWorks thing remains to be seen.

Sean Phillips wrote: